2026 ASEE Annual Conference & Exposition

Teamwork has become a cornerstone of higher education, preparing students to navigate complex, interdisciplinary, and global problems. Yet students frequently struggle to collaborate effectively due to unequal participation, miscommunication, and differing expectations about roles and responsibilities. Research shows that these challenges are often rooted not only in technical skill gaps but also in limited metacognitive capacity. Metacognition refers to students’ ability to monitor, evaluate, and regulate their collaborative processes. Traditional mentoring can scaffold these practices, but in large classrooms it is rarely feasible to provide individualized, ongoing support. This paper introduces EMPA, an AI-based collaborative partner designed to extend mentoring capacity by engaging learners in cycles of metacognitive knowledge and regulation. EMPA interacts with students through case studies, videos, and dialogic prompts, encouraging them to surface assumptions, examine diverse perspectives, monitor group dynamics, and evaluate strategies for future collaboration. The system integrates global competence dimensions such as differing approaches to communication, decision-making, and time management, while situating them within broader models of self-regulated learning. In this way, EMPA positions awareness of multiple perspectives as part of metacognitive knowledge and links reflective practice with the regulation of collaborative behaviors. A pilot implementation with undergraduate students demonstrated how EMPA mediated both metacognitive awareness and regulation. Learners initially described teamwork breakdowns in personal terms but, through EMPA’s scaffolds, reframed them as differences in communication or work orientation. They sought clarification of concepts, revised initial assumptions, and articulated strategies to foster more balanced collaboration. Importantly, students connected these insights to their prior teamwork experiences, identifying regulatory actions such as clarifying deadlines, distributing roles, or setting shared expectations that they planned to adopt in future projects. These findings highlight the potential of AI to serve as a co-regulator of metacognition. Rather than functioning solely as a feedback mechanism, EMPA engaged learners across the full cycle of awareness, monitoring, adjustment, and evaluation. By doing so, it bridged classroom simulations with authentic teamwork challenges, demonstrating how AI can support not only cognitive and technical skills but also the adaptive, globally relevant capacities essential for collaboration. We argue that integrating AI-mediated metacognition into teamwork education offers a scalable pathway to equip students with the reflective and regulatory skills needed for effective participation in diverse teams.